Plotting Plays, Python Data (Standardized Direction and Location)

0.1 Topic

This file

0.2 Set Up

Packages/Files

library(tidyverse)
library(magrittr)
library(ggforce)

Load Data

loadData <- function(fileName = "train.csv", ...) {
  file = paste("../input", fileName, sep="/")
  data.table::fread(file, ...) %>% 
    as_tibble()
}
df = loadData("trainClean_py.csv")
df %<>%
  mutate(
    X_end = S*cos((90-Dir)*pi/180) + X, 
    Y_end = S*sin((90-Dir)*pi/180) + Y
  )
dfFeatures = loadData("features_py.csv")

1 Plot code from rescale file

Taken from https://www.kaggle.com/statsbymichaellopez/nfl-tracking-wrangling-voronoi-and-sonars#Dealing-with-player-tracking-data

options(repr.plot.width=10, repr.plot.height=6)
samp_play <- "20170910001102"
plotPlay = function(dfPlay, features=NULL, run=F) {
  p = dfPlay %>% 
    ggplot(aes(X, Y, color = OnOffense))  + 
    geom_point(size = 2) + 
    geom_segment(aes(x = X, y = Y, xend = X_end, yend = Y_end), 
                 arrow = arrow(length = unit(.5,"cm"))) + 
    geom_point(data = filter(dfPlay, NflId == NflIdRusher), 
               pch=21, size = 1.5, fill = "black") +
    # geom_point(data = filter(dfPlay, Position == "QB"), size = 2, fill = "blue") +
    scale_colour_brewer(palette = "Set2")+ 
    scale_fill_brewer(palette = "Set2")+ 
    # geom_vline(aes(xintercept=0), color="grey") +
    # geom_vline(aes(xintercept=10), color="grey") +
    geom_vline(aes(xintercept = LineOfScrimmage), colour = "black", lty = 2) + 
    coord_cartesian(ylim=c(0,160/3), expand=F) + #xlim = c(-10,110), 
    labs(
      x = "Distance from offensive team's own end zone", 
      y = "Y", 
      title = paste0("PlayId ", pull(dfPlay, PlayId)), 
      subtitle = paste("Yards", dfPlay$Yards, "; ", "Down", dfPlay$Down)
    ) + 
    theme_bw(14) + 
    theme(panel.grid.minor = element_blank(), panel.grid.major.y =element_blank())
  if (run && !is.null(features)) {
    dfC = circleDF(c(features$LineOfScrimmage, features$Rusher_Gap_Center),
                   features$Rusher_Gap_Radius)
    p + 
      geom_path(data = dfC, aes(x, y), color="gold") +
      labs(subtitle = paste("Yards", dfPlay$Yards, "; ", "Down", dfPlay$Down, "; ",
                            "NPlayers", features$Rusher_Gap_NPlayers, "; ",
                            "OpenSize", round(features$Rusher_Gap_OpenSize, 2)))
  } else {
    p
  }
}
circleDF = function(center = c(0,0), r = 1, npoints = 100){
    tt <- seq(0, 2*pi, length.out = npoints)
    xx <- center[1] + r * cos(tt)
    yy <- center[2] + r * sin(tt)
    return(data.frame(x = xx, y = yy))
}
dfPlay = filter(df, PlayId == samp_play)
features = filter(dfFeatures, PlayId == samp_play)
plotPlay(dfPlay, features, run=T)

2 Looking at Plays

First, categorize runs: long, medium, short, bad

ggplot(df) +
  geom_density(aes(Yards))

summary(df$Yards)

   Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
-14.000   1.000   3.000   4.212   6.000  99.000 

df %<>%
  mutate(YardsCategory = cut(Yards, breaks = c(-Inf, 0, 3, 6, Inf), 
             labels=c("bad", "short", "med", "long")))
table(df$YardsCategory)

   bad  short    med   long 
103532 184888 117106 104236 

PlayId_long = df %>% 
  filter(YardsCategory == "long") %>% 
  pull(PlayId)
PlayId_med = df %>% 
  filter(YardsCategory == "med") %>% 
  pull(PlayId)
PlayId_short = df %>% 
  filter(YardsCategory == "short") %>% 
  pull(PlayId)
PlayId_bad = df %>% 
  filter(YardsCategory == "bad") %>% 
  pull(PlayId)

Plot several runs from vector of ids

plotPlays = function(ids, n=5) {
  for (index in sample.int(length(ids), min(n, length(ids)))) {
    id = ids[index]
    dfPlay = filter(df, PlayId == id)
    features = filter(dfFeatures, PlayId == id)
    p = plotPlay(dfPlay, features, run=T)
    print(p)
  }
}

2.1 Long Runs

plotPlays(PlayId_long)

"20181007032152"

[1] "20181007032152"

strange play

plotPlays("20181007032152")

filter(df, PlayId=="20181007032152")

# A tibble: 22 x 38
   PlayId GameId     X     Y     S     A   Dis Orientation   Dir  NflId
   <int6>  <int> <dbl> <dbl> <dbl> <dbl> <dbl>       <dbl> <dbl>  <int>
 1 20181… 2.02e9  18.5  43.1  1.99  2.91  0.18       -82.9  65.7 4.36e3
 2 20181… 2.02e9  32.0  19.2  2.72  1.96  0.27       -58.0  56.8 2.50e6
 3 20181… 2.02e9  15.8  31.5  2.86  3.76  0.27       245.  -44.9 2.51e6
 4 20181… 2.02e9  19.9  28.8  3.54  3.02  0.33       -80.5 -35.2 2.51e6
 5 20181… 2.02e9  23.0  31.1  4.18  3.25  0.4        -84.6 -62.1 2.54e6
 6 20181… 2.02e9  20.8  24.0  2.48  1.85  0.24       -84.1 -27.8 2.54e6
 7 20181… 2.02e9  16.6  28.9  1.85  2.72  0.18       -86.3 -78.1 2.55e6
 8 20181… 2.02e9  19.8  17.3  3.02  3.67  0.28       269.  201.  2.56e6
 9 20181… 2.02e9  16.7  26.5  2.84  2.48  0.31       261.  -24.5 2.56e6
10 20181… 2.02e9  16.5  22.0  2.24  0.41  0.22       -87.8 -31.9 2.56e6
# … with 12 more rows, and 28 more variables: Quarter <int>,
#   PossessionTeam <int>, Down <int>, Distance <int>, FieldPosition <int>,
#   NflIdRusher <int>, OffenseFormation <int>, OffensePersonnel <int>,
#   DefendersInTheBox <dbl>, DefensePersonnel <int>, PlayDirection <chr>,
#   TimeHandoff <chr>, TimeSnap <chr>, Yards <int>, PlayerHeight <int>,
#   PlayerWeight <int>, Position <chr>, HomeTeamAbbr <int>,
#   VisitorTeamAbbr <int>, Week <int>, StadiumType <int>, Turf <int>,
#   GameWeather <int>, LineOfScrimmage <int>, OnOffense <lgl>,
#   X_end <dbl>, Y_end <dbl>, YardsCategory <fct>

2.2 Medium Runs

plotPlays(PlayId_med)

2.3 Short Runs

plotPlays(PlayId_short)

2.4 Bad Runs

plotPlays(PlayId_bad)

2.5 To Edge

edgePlays = dfFeatures %>% 
  filter(Rusher_Gap_ToEdge == 1) %>% 
  pull(PlayId)
plotPlays(edgePlays)